PROJECTQuantitative Data Analysis

Project Navigation
  • Project Home
  • Inference
  • Impact Analysis
  • Bias
  • Experiments
  • Paired Testing
  • Quasi-experimental Methods
  • Difference-in-Difference and Panel Methods
  • Instrumental Variables
  • Propensity Score Matching
  • Regression Discontinuity
  • Regression Techniques
  • Generalized Linear Model
  • Linear Regression
  • Logit and Probit Regression
  • Segregation Measures
  • Inequality Measures
  • Decomposition Methods
  • Descriptive Data Analysis
  • Microsimulation
  • The Dynamic Simulation of Income Model DYNASIM
  • The Health Insurance Policy Simulation Model HIPSM
  • The Model of Income in the Near Term (MINT)
  • The Transfer Income Model TRIM
  • The Tax Policy Center Microsimulation Model
  • Performance Measurement and Management

  • Introduction

    The Health Insurance Policy Simulation Model (HIPSM) is a detailed microsimulation model of the health care system designed to estimate the cost and coverage effects of proposed health care policy options. Changes to individual or employer decisions in one insurance market interact with decisions in other markets. The model is iterative, in that policy changes are simulated, information from that simulation is fed back into the model, and the model continues to iterate until there is little to no change in coverage and costs between iterations, i.e., until the model reaches equilibrium. HIPSM is designed for quick-turnaround analysis of policy proposals. It can be rapidly adapted to analyze a wide variety of new scenarios—from novel health insurance offerings and strategies for increasing affordability to state-specific proposals—and can describe the effects of a policy option over several years.

    HIPSM is based on two years of the American Community Survey, which provide a representative sample of families that is large enough for us to produce estimates for individual states and smaller regions such as cities. The model is designed to incorporate timely, real-world data to the extent that they are available. We regularly update the model to reflect published Medicaid and marketplace enrollment and costs in each state. The enrollment experience in each state under current law affects how the model simulates policy alternatives.

    Results from HIPSM simulations have been favorably compared to actual policy outcomes and compared to other respected microsimulation models, as assessed by outside experts.1 Findings from the model were cited in the majority opinion in the Supreme Court case King v. Burwell, and in many amicus briefs submitted to the court in that case. Findings from HIPSM have been broadly cited in top tier media, including the New York Times, Washington Post, Wall Street Journal, Vox, CNN, and Los Angeles Times. Results from HIPSM have been displayed on the floor of the US Senate during debate, and are widely distributed among legislative staff.

    Policy Analyses Using HIPSM

    The Urban Institute team has produced a broad array of state specific modeling (often analyses of each of the 50 states plus the District of Columbia) of health policies of national interest. These analyses are generally funded by philanthropic foundations, such as the Robert Wood Johnson Foundation, The Commonwealth Fund, and Arnold Ventures, or government funders, such as the Medicaid and CHIP Payment and Access Commission (MACPAC). We have also provided detailed technical assistance to several state governments, most recently New Mexico, New York, and Virginia.  

    Important recent analyses using HIPSM include the following:

    HIPSM Output

    Outputs of the model can be designed to meet the specific needs of funders, but past outputs include comparisons of the situation under current law versus under a policy change, the impact on state and federal spending, and detailed characteristics of those who would gain or lose coverage. Specifically, estimates for results tables frequently include, but are not limited to, the following:

    • Program eligibility. Medicaid, CHIP, BHP, marketplace premium tax credits (PTCs), cost sharing reductions (CSRs), exemptions from the individual mandate.
    • Type of coverage. Employer, marketplace (PTCs and CSRs, PTCs only, full pay), other nongroup, BHP, Medicaid (children, children with disabilities, nonparents, parents, adults with disabilities), CHIP, other public (including Medicare), and uninsured
    • Socioeconomic characteristics, such as income group, age, race/ethnicity (including Asians/Pacific Islanders and American Indians/Alaska Natives), educational attainment, employment status, family structure, immigration status, English proficiency, and language spoken at home.
    • Tabulations by sub-state region. Regions must be defined by Census Department 2010 Public Use Microdata Areas (https://www.census.gov/geo/reference/puma.html).
    • Medicaid-related costs (per capita or total). State and federal shares.
    • BHP-related costs (per capita or total). Out-of-pocket (OOP) premiums, OOP cost sharing, federal BHP payments, and BHP costs to the state.
    • Marketplace QHP costs (per capita or total). OOP premiums, OOP cost sharing, federal PTCs, federal CSRs, and total premiums. Also, additional costs if the state were to supplement federal PTCs or CSRs.
    • Other costs. Uncompensated care, employer premium contributions, and total premiums for employer health coverage.
    • Health cost risk scores for any group of nonelderly persons.
    • Health care costs may be separated into hospital, physician, prescription drugs, and other.

    More about the Model

    Read a complete description of HIPSM’s methodology. A summary of the construction of HIPSM’s baseline (current law scenario) is as follows:

    • As the core data, we use the US Census Bureau’s American Community Survey (ACS) from 2012 and 2013, which we combine to increase sample size (over 6 million observations). The combined file was reweighted to reflect the distribution of demographic, economic, and health coverage characteristics of the 2013 ACS. We use pre-ACA data as a model baseline because later ACS data reflect the ACA’s transitional period which differs both from pre-ACA conditions and from the eventual full impact of the ACA.  
    • Each year, the model is calibrated to reproduce the latest available Medicaid and marketplace enrollment numbers in each state under the ACA.
    • Population weights for future years were derived from the Urban Institute’s Mapping America’s Futures program. These projections match Census Bureau population projections nationally, but include much more detail, particularly at the state level;
    • Using the Medical Expenditure Panel Survey-Household Component (MEPS-HC) and other data sources, we estimate health care expenditures for each individual in the data set in each possible coverage status, including out-of-pocket spending, spending covered by insurance, Medicaid/CHIP spending, and uncompensated care for the uninsured;
    • We impute offers of employer-sponsored insurance, immigration status, and eligibility for Medicaid, the Children’s Health Insurance Program (CHIP), and subsidized qualified health plan (QHP) coverage; and
    • We group together workers with the same employment characteristics, such as firm size and industry, into simulated firms. The distribution of these firms matches the characteristics of employers in each Census Division provided in the US Statistics of Business.

    The general flow of a HIPSM simulation is as follows:

    • The model constructs available insurance packages and computes premiums based on current enrollment and insurance market regulations;
    • Simulated employers choose whether or not to offer coverage and whether to offer coverage inside or outside the marketplace (if applicable);
    • Individuals and families choose from among the coverage options available to them: employer-sponsored insurance, nongroup insurance, health insurance marketplaces (if applicable), Medicaid/CHIP, or uninsured;
    • Employer, individual, and family decisions are calibrated so that overall behavior is consistent with results from the health economics literature; and
    • Premiums are updated based on the new enrollment decisions. The cycle is repeated until equilibrium—in other words, until there is little change between successive iterations of the model.

    HIPSM is unique among microsimulation models of health coverage and costs because individual and family decisions combine the two most common types of microsimulation decisionmaking, elasticity and expected utility. Other models use one or the other:

    • Elasticity models. Estimate premium and cost changes and apply elasticities from the literature.  However, health coverage options substantially different from existing ones do not naturally fit in this framework, so assumptions must be made based on existing options.
    • Expected utility models. Define an expected utility function that takes into account expected out-of-pocket spending, health needs, risk of high health costs, and income. Each unit chooses the option with the highest expected utility.  Allows for the evaluation of novel policies in the same framework. However, in order to replicate a health coverage baseline, actual data on health coverage choices must be reconciled with a theoretical utility function.
    • HIPSM. Decision-making follows an expected-utility framework, but we add a latent preference term for each observation that represents factors involved in their choice that we could not capture.  These terms are set so that each observation makes the choice it reported, and the distribution of latent preference terms is set so that the model replicates elasticity targets from the literature if premiums rise or fall. This approach makes it easier to simulate novel policies in a consistent way, while also calibrating the model to a wide range of real-world data such as Medicaid and marketplace enrollment.

    In addition, HIPSM alone is based on a core of the ACS, allowing for state specific findings. Read more about the HIPSM methodology.

    Research Methods Microsimulation modeling Data analysis Quantitative data analysis Research methods and data analytics